基于属性分组的子空间聚类算法研究

Subspace clustering algorithm based on attribute grouping

针对分类数据,基于属性分组技术和多目标聚类质量函数,提出一种子空间聚类算法.该算法采用属性分组技术,将高相关属性划分到同属性组中,利用同组属性相关性度量属性权重值,构建属性软子空间;采用基于多目标的聚类质量函数,判断整体聚类效果,通过迭代优化簇集结构,达到最佳的数据划分状态.在人工合成数据集和UCI数据集上,实验验证了该算法的正确性、高效性和可靠性.

A subspace clustering algorithm was proposed based on attribute grouping technology and multi-objective clustering quality function for categorical data.The algorithm used attribute grouping technology to divide highly correlated attributes into the same attribute group,and used the correlation of attributes in the same group to measure the attribute weight value to build the attribute soft subspace;The cluster quality function based on multiple objectives was adopted to judge the overall clustering effect,and the cluster structure was optimized iteratively to achieve the best data partition state.The correctness,efficiency,and reliability of the algorithm were verified by experiments on synthetic datasets and UCI datasets.